# lavaan package for Structural Equation Model in R

lavaan is a package for running a “Structural Equation Models” or SEM in R. The script of lavaan is similar to Mplus, which is very comprehensive and compact. The lavaan package is developed by Yves Rosseel.

The following is my lavaan script I used in my presentation at R useR! 2011.

##-------------------------------------------------## ## Measuring Transaction Cost in Supply Chains ## ## Pairach Piboonrungroj ## ## R useR conference August 2011 ## ##-------------------------------------------------## # 1. Load data hoteldata <- read.csv("https://www.dropbox.com/s/fzh040xkhofozjy/cleandata.csv") # 2. Install Package install.packages("lavaan") # 3. Load Package library(lavaan) #5. Structural Model TC.Model <- ' # latent variable definitions cost =~ TC1 + TC2 + TC3 +TC6 + TC7 +TC11 + TC13 asset =~ AS1 + AS2 uncertainty =~ UN1 + UN2 # regression cost ~ uncertainty + asset # resident (co)variance TC1 ~~ TC3 TC6 ~~ TC13 TC11 ~~ TC13 TC13 ~~ AS1 ' fitTC <- sem(TC.Model, data = hoteldata) summary(fitTC, standardized = TRUE, fit.measures=TRUE) # Coefficients # coef(fitTC) #More indices # Fit indices e.g. CFI fit.indices <- inspect(fitTC, what = "fit") fit.indices["cfi"] # Calling Modification Indices # inspect(fitTC, what = "mi")

See more details about the model and the comparison with other R packages and software

8 Comments
Post a comment

‘t would be good to have some data to try this script out, can you post either the csv or the structure of the csv file?

Hi, thanks a lot for your interest.

I have put a public of the data file in the script already. Then one can run the script straight away. Please let me know any bug or problem you find.

Hi I tried to use that script comparing with the amos results. The regression and correlation are almost similar however the model fit index especially the chi p-value cannot provide result if the df is 1. The amos recognize the df as 1 while this r-code gives 0 df. Any way to improve the model fit script ?

Degrees of freedom I’ve got Degrees of freedom of 37 using the R code above.

Have you included the whole code?

Thanks for the reply. Yes sir i included the codes above using my own data. By comparing the results using my data, the Amos gives DF =1 while this R-script gives DF=0. With that, the chi square p-value don’t provide results for model fit. Another thing is that the CFI, TLI values tends to round-off. Meanwhile the RMSEA values seems to have problem.

Additionally, can we use this codes to compute sem model without latent variable just purely the exogenous and endogeonus something like causal modelling?